AI and Jevons Paradox
Amazon’s AI CEO, Andy Jassy, made a splash—in a bad way—last year when he announced that Artificial Intelligence (AI) will replace jobs there. So much so that in December 2025, over 1,000 employees signed a letter complaining about AI, in part to save their jobs. (Personally, those would be the first to go, in my opinion.)
But the fear is real.
However, in a follow up discussion earlier in 2026, Jassy clarified:
I do believe that a lot of the jobs that we’ve thrown human beings at for the last 20 or 30 years, you won’t need as many human beings doing those same jobs […] I also think there are going to be other jobs created, and that has always happened in every technology shift.
Will AI take jobs? Yes.
Will AI make jobs? Yes.
These two questions are not a contradiction. They may seem like it, but only if you don’t understand Jevons’ paradox.
Let me explain.
Jevons Paradox
In 1865, English economist William Stanley Jevons, noted that the increased efficiency of the steam engine actually increased coal use. There was concern at the time that England was going to run out of coal reserves and arguments were made trying to improve efficiency in order to decrease coal usage. Jevons argued that this was incorrect. He noted in his 1865 work, The Coal Question, that Watt’s improvements in the efficiency of the steam engine made coal more economical for use. This increased the industrial applications the steam engine could be used and thus, led to an increase in coal consumption.
The reality of this is tied to the law of demand which notes that as the price of a good or service decreases then the demand for the good or service increases. And in the case of coal, England did not run out. They had to go deeper to mine coal.
While it is not a perfect reality with AI, other technological changes have reshaped the way economies work. I am sure the introduction of the motor car and the price efficiency of the Model T made farrier’s nervous. No longer would they be needed to care for horse’s hooves. But they could retrain and become Goodyear tire salesmen instead.
The Reality
Technological revolutions increase efficiency or application in one thing that leads to increased productivity or other advancements. Yet, through them all the means and method of work have shifted and humans adapt. One article makes the case that a technology revolution is followed closely by a skills revolution to adapt to the new technology.
This misses the mark with AI.
Box CEO Aaron Levie recognizes this reality in a recent podcast. He states:
If you or I go and vibe-code something, we think we’ve replaced the engineer, replaced the accountant, replaced the lawyer,” Levie told me. “But then you actually look — that was the first 80% of the job. The extra 20%, it turns out, is all the value creation of that profession. All the expertise and domain knowledge is in that last 20%, not the text that got generated.
It is the “last mile” of the human effort that is important because humans have real intelligence (RI). We inately learn to recognize patterns and process information and fill in the gaps. We don’t have to be programmed or taught thousands of times.
If I list the following numbers
1 1 2 3 5 8 13
and ask you to complete the sequence you can probably tell me the next number is 21 and then 34 and so on. And you can recognize it as the Fibonacci sequence.
Humans are good at patterns and analysis and the real benefit of AI is that it will gather all the data and information[i] and humans with RI can have more time doing what humans do best: analyzing, processing, pattern recognition, etc.
That is why you should not be afraid of AI.
How does one differential themselves in this brave new world of AI?
Use your RI.
[i] For the purposes of this discussion, we are going to assume that the data and information that the AI system gathers for your analysis, etc. is correct and error free. There is concern that the LLMs are still hallucinating. That is a key component of how humans use their RI to differentiate themselves in the AI market. The subject matter expert should review any AI generated outputs to ensure that they make sense. For one discussion, see https://blogs.library.duke.edu/blog/2026/01/05/its-2026-why-are-llms-still-hallucinating/.
Kevin Robinson, CISSP, DDN.QTE, Associate C|CISO, is Head of Cybersecurity Services for The Commonwealth Group. He has a 20 year career in cybersecurity, risk assessment, intelligence and counterintelligence. His previous employers include Thornburg Investment Management, Los Alamos National Laboratory, L3Harris, and the Central Intelligence Agency.